Investigation of cavity filling in profile ring rolling process for T-shape ring

Abstract

In this paper, key parameters affecting the cavity filling in single and double T-shape profile rings are comprehensively investigated via numerical and experimental analysis. A three-dimensional finite element model was developed in Abaqus\Explicit to assess the influence of crucial ring rolling process parameters, including feed speed, main roll rotational velocity, the existence and the absence of axial rolls on the cavity filling of single and T-shape rings and the main roll torque. Besides, a ring rolling machine was built to conduct practical experiments and validate the numerical evaluation, while for the first time, the role of the axial roll and the main roll torque on the quality of the cavity filling is experimentally evaluated. Power requirements and the final ring profile geometry were obtained by the simulation method, and the results were confirmed by the experiments. The results showed that axial rollers significantly reduced the cavity filling rate, and in contrast, the effect of mandrel feed speed and the main roll rotational velocity was much lower. Also, the axial forces were considerably less than the radial forces. However, the rolling operation was done in both radial and axial directions. The existence of axial rolls had an intensive effect on the process’ required power, as a result the main roll torque increased more than three times in case of applying axial rolls, compared with not considering them. Severe effects of axial rollers on increasing force and decreasing cavity filling rate can be attributed to frictional forces between the ring and axial rolls, restricted ring motion, which has to be compensated by a higher torque of the main roll. When the axial rolls are used, the material flow in the ring’s height direction is restricted. Therefore, the material cannot move easily to form the profile. All experimental and simulation results, including mandrel force, cavity filling, and ring profile geometry, were in good agreement, and in all cases, the simulation error was less than 10%.